Stochastic Transceiver Optimization in Multi-Tags Symbiotic Radio Systems

Xihan Chen, Hei Victor Cheng, Kaiming Shen, An Liu, Min-Jian Zhao

TL;DR

This paper introduces a novel framework for optimizing "Symbiotic Radio" (SR) systems, where passive IoT tags transmit data by backscattering existing active radio signals. The authors develop a Batch Stochastic Parallel Decomposition (BSPD) algorithm that solves the complex, non-convex math required to manage interference in multi-tag environments, significantly boosting throughput and reliability for both the primary transmission and the passive sensors.

What Problem Does This Paper Address?

The Internet of Things (IoT) faces a dual crisis: battery life and spectrum scarcity. Symbiotic Radio (SR) is a promising solution where passive "Tags" (like smart sensors) don't generate their own radio waves; instead, they "hitchhike" on the signals of an active primary transmitter (like a WiFi router).

However, moving from a single-tag to a multi-tag system creates a mathematical nightmare. The primary signal interferes with the tags, and the tags interfere with each other. Furthermore, these tags are too simple to perform complex tasks like adaptive coding. Existing optimization methods often require real-time knowledge of every tag's symbol—a feat that is practically impossible due to signaling overhead. The challenge is to optimize the system using only statistical information while guaranteeing a minimum Quality of Service (SINR) for every device.

Methods

The authors formulate the problem as a General Network Utility Maximization Problem (GUMP). To solve it, they employ several sophisticated techniques:

  • Barrier Method: Instead of treating minimum SINR requirements as "hard" constraints (which are difficult to optimize), they incorporate them into the objective function using a logarithmic barrier. This penalizes the system as it approaches the SINR threshold, ensuring QoS with high accuracy.
  • Fractional Programming (FP): The SINR equations are ratios of non-convex functions. The authors use FP techniques to transform these into a more manageable "surrogate" function.
  • BSPD Algorithm: The core contribution is the Batch Stochastic Parallel Decomposition algorithm. It processes data in "mini-batches" (similar to how deep learning models are trained) to estimate the expected values of the random wireless channels.
  • Decoupling: The algorithm is designed so that the optimization of the transmit beamforming (at the router) and the receive beamforming (at the receiver) can be solved independently and in parallel, greatly reducing computational lag.

Datasets

The paper relies on Monte Carlo simulations rather than a static dataset. The authors modeled a single-cell SR system with:

  • Primary Transmitter (PT): 64 antennas.
  • Primary Receiver (PR): 16 antennas.
  • Tags: 4 single-antenna passive tags randomly distributed in a specific region.
  • Channel Models: Independent Rayleigh fading with distance-dependent path loss (modeled at 900 MHz).
  • Baselines: The proposed scheme was compared against three standard benchmarks: EVD-based beamforming, MMSE receivers, and a maximum channel gain scheduling criterion (Baseline 3).

Key Insights and Ideas

  • Mutualism, not Parasitism: While tags are "parasitic" in that they use the primary signal, the relationship is truly symbiotic. The backscattered signals from the tags actually provide additional "multipath" components that the primary receiver can use to improve the main signal's decoding.
  • Statistical over Instantaneous: By optimizing based on the statistics of the tag symbols rather than their real-time values, the authors bypass the "huge signaling overhead" that usually kills the efficiency of backscatter systems.
  • Structure Preservation: The "surrogate function" isn't just a simplification; it's carefully tailored to preserve the underlying structure of the SINR problem, which is why it converges to a stationary solution much faster than generic optimization tools.

Key Results

  • Convergence: The BSPD algorithm demonstrated rapid convergence to a stationary solution within approximately 30–40 iterations.
  • Throughput Gains: The proposed scheme achieved a DL transmission rate of 6.167 bps/Hz, significantly outperforming the best baseline (5.342 bps/Hz).
  • Reliability: The "Worst-case Tag BER" (Bit Error Rate) was reduced to $6.129 \times 10^{-3}$, while all baseline schemes suffered from much higher error rates (ranging from $10^{-1}$ to $0.5$).
  • Scalability: The performance gap between the proposed scheme and baselines widened as the number of transmit antennas increased, showing that the algorithm effectively leverages large antenna arrays (MIMO).

Limitations and Future Work

  • Perfect Interference Cancellation: The primary model assumes perfect Direct Link Interference (DLI) cancellation. While the authors mention the algorithm can be extended to imperfect DLI (Remark 1), they do not provide detailed simulations for those "messy" real-world scenarios.
  • Static Channels: The study assumes the channel is static within a block (block-fading). Future work could explore high-mobility environments where the channel changes rapidly.
  • Modulation Complexity: The tags are limited to On-Off Keying (OOK). Exploring higher-order modulation for passive tags within this stochastic framework remains an open challenge.

Summary updated

Abstract

Symbiotic radio (SR) is emerging as a spectrum-and energy-efficient communication paradigm for future passive Internet of Things (IoT), where some single-antenna backscatter devices, referred to as Tags, are parasitic in an active primary transmission. The primary transceiver is designed to assist both direct-link (DL) and backscatter-link (BL) communication. In multi-Tags SR systems, the transceiver designs become much more complicated due to the presence of DL and inter-Tag interference, which further poses new challenges to the availability and reliability of DL and BL transmission. To overcome these challenges, we formulate the stochastic optimization of transceiver design as the general network utility maximization problem (GUMP). The resultant problem is a stochastic multiple-ratio fractional nonconvex problem, and consequently challenging to solve. By leveraging some fractional programming techniques, we tailor a surrogate function with the specific structure and subsequently develop a batch stochastic parallel decomposition (BSPD) algorithm, which is shown to converge to stationary solutions of the GNUMP. The simulation results verify the effectiveness of the proposed algorithm by numerical examples in terms of the achieved system throughput.
BibTeX
@article{chen2020stochastic,
  author = {Xihan Chen and Hei Victor Cheng and Kaiming Shen and An Liu and Min-Jian Zhao},
  title = {{Stochastic Transceiver Optimization in Multi-Tags Symbiotic Radio Systems}},
  journal = {IEEE Internet of Things Journal},
  year = {2020},
  url = {https://openreview.net/forum?id=pGSWUwG-WBH},
}

Loading related content...

Figure